ovarian cancer stem cells and macrophages reciprocally ...educated by tumor-derived factors into an...

RESEARCH ARTICLE Open Access

Ovarian cancer stem cells andmacrophages reciprocally interact throughthe WNT pathway to promote pro-tumoraland malignant phenotypes in 3Dengineered microenvironmentsShreya Raghavan1, Pooja Mehta1, Yuying Xie6, Yu L. Lei4,5 and Geeta Mehta1,2,3,5*

Abstract

Background: Innate immune cells such as macrophages are abundantly present within malignant ascites, wherethey share the microenvironment with ovarian cancer stem cells (CSC).

Methods: To mimic this malignant ascites microenvironment, we created a hanging-drop hetero-spheroid modelto bring CSCs and macrophages in close association. Within these hetero-spheroids, CD68+ macrophages (derivedfrom U937 or peripheral blood monocytes) make up ~ 20% of the population, while the rest are ovarian cancercells and ovarian cancer stem cells (derived from the high grade serous ovarian cancer cell line, OVCAR3).

Results: Our results indicate that CSCs drive the upregulation of M2 macrophage marker CD206 within hetero-spheroids, compared to bulk ovarian cancer cells, implying an inherently more immuno-suppressive program.Moreover, an increased maintenance of elevated aldehyde dehydrogenase (ALDH) activity is noted within hetero-spheroids that include pre-polarized CD206+ M2 macrophages, implying a reciprocal interaction that drives pro-tumoral activation as well as CSC self-renewal. Consistent with enriched CSCs, we also observe increased levels of pro-tumoral IL-10 and IL-6 cytokines in the CSC/M2-macrophage hetero-spheroids. CSC/M2-macrophage hetero-spheroidsare also less sensitive to the chemotherapeutic agent carboplatin and are subsequently more invasive in transwellassays. Using inhibitors of WNT secretion in both CSCs and macrophages, we found that CSC-derived WNT ligandsdrove CD206+ M2 macrophage activation, and that, conversely, macrophage-derived WNT ligands enriched ALDH+

cells within the CSC compartment of hetero-spheroids. Upon examination of specific WNT ligand expression within themonocyte-derived macrophage system, we observed a significant elevation in gene expression for WNT5B. In CSCs co-cultured with macrophages within hetero-spheroids, increases in several WNT ligands were observed, and this increasewas significantly inhibited when WNT5B was knocked down in macrophages.

Conclusions: Our data implies that macrophage- initiated WNT signaling could play a significant role in themaintenance of stemness, and the resulting phenotypes of chemoresistance and invasiveness. Our results indicateparacrine WNT activation during CSC/M2 macrophages interaction constitutes a positive feedback loop that likelycontributes to the more aggressive phenotype, which makes the WNT pathway a potential target to reduce the CSCand M2 macrophage compartments in the tumor microenvironment.

© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

* Correspondence: [email protected] of Materials Science and Engineering, 2800 Plymouth Rd,Building 28, Room 3044W, Ann Arbor, MI 48109, USA2Department of Biomedical Engineering, 2800 Plymouth Rd, Building 28,Room 3044W, Ann Arbor, MI 48109, USAFull list of author information is available at the end of the article

Raghavan et al. Journal for ImmunoTherapy of Cancer (2019) 7:190 https://doi.org/10.1186/s40425-019-0666-1

on May 13, 2021 by guest. P

rotected by copyright.http://jitc.bm

j.com/

J Imm

unother Cancer: first published as 10.1186/s40425-019-0666-1 on 19 July 2019. D

ownloaded from

http://crossmark.crossref.org/dialog/?doi=10.1186/s40425-019-0666-1&domain=pdf

http://orcid.org/0000-0001-5967-6425

http://creativecommons.org/licenses/by/4.0/

http://creativecommons.org/publicdomain/zero/1.0/

mailto:[email protected]

http://jitc.bmj.com/

IntroductionLate stage epithelial ovarian cancer presents frequentlywith peritoneal carcinomatosis, and is associated with theformation of malignant ascites. Exfoliated ovarian cancercells exist as spheroids within the ascites microenviron-ment, and there is also an enrichment of ovarian cancerstem cells (CSCs) within the peritoneal fluid [1–3]. Withinthe malignant ascites, CSCs interact with a variety of hostcells including different immune subsets in the presenceof a complex cytokine/chemokine network, ultimatelyleading to trans-coelomic metastasis [4–6]. Also found inabundance within the peritoneal ascites fluid are macro-phages, which are generally believed to be polarized andeducated by tumor-derived factors into an M2-resemblingpro-tumoral phenotype. In fact, ovarian cancer cells inter-act with macrophages in anchorage independent condi-tions and grow as spheroids within the malignant ascites,which can result in tumor metastasis even during the earlystage of peritoneal dissemination [7].In the ovarian cancer microenvironment, macrophages

are activated into a tumor-associated macrophage (M2-like,“M2”, alternatively activated) phenotype primarily througheducation by tumor-derived cytokines, chemokines andother tumor cell-derived factors [8, 9]. Tumor-associatedmacrophages within malignant ascites have a bipolar ex-pression spectrum, ranging from M1-like to M2-like phe-notypes. However, M2-like macrophages are the primarypro-tumoral phenotype in the peritoneal cavity. In fact, ahigh ratio of M1/M2 macrophages is associated with an im-proved prognosis in ovarian cancer, whereas lower M1/M2ratio is indicative of a poor prognosis [10–12].Little is known about the specific role of cancer stem

cells in macrophage polarization and activation. Even lessis known about the reciprocal interactions between CSCsand macrophages. Given the enriched presence of macro-phages and CSCs within the malignant ascites, their inter-action may be critical for regulating the progression anddrug response of ovarian cancer. Therefore, in this study,we used a previously established in vitro hanging dropspheroid model [13–15] to dissect the reciprocal interac-tions between the CSCs in tumor spheroids and macro-phages. The hanging drop spheroid model allows theformation of stable spheroids in a non-adherent 3D invitro environment, similar to the aggregation of ovariancancer cells floating within the malignant ascites inanchorage-independent conditions. We previously dem-onstrated that the hanging drop spheroid model maintainsCSCs derived from primary patient samples with high fi-delity, and preserves responses to chemotherapeuticagents similar to mouse xenograft models [13].The importance and abundance of WNT-signaling have

been demonstrated in ovarian development, tumorigenesisand stem cell maintenance [16, 17]. In the tumor immunemicroenvironment, activated WNT/β-catenin signaling

can suppress the recruitment of dendritic cells, therebylimiting T-cell priming, and intra-tumoral T-cell accumu-lation [18]. WNT signaling is also heavily involved in theactivation of macrophages [19, 20]. Importantly, paracrineWNT signaling loops between M2-like macrophages andtumor cells contribute to tumorigenesis and invasiveness[21, 22].We hypothesized that any trophic interactions be-

tween CSCs and macrophages may involve a WNT-dependent pathway. Therefore, using the hanging dropspheroid model, we sought to understand pro-tumoralmacrophage activation in response to CSCs, and changesin the CSC compartment itself in response to activatedmacrophages. We evaluated the WNT pathway in CSC-macrophage interactions, and whether that corre-sponded to functional changes in chemoresistance or in-vasion of CSC spheroids. Insight into WNT involvementin CSC-macrophage interactions could provide new tar-getable avenues to reduce CSC-burden in ovarian cancer,thereby limiting metastatic and recurrent disease.

Materials and methodsMaterialsCell lines were purchased from ATCC (Manassas, VA).Peripheral blood mononuclear cells (PBMCs) were puri-fied from buffy coats from healthy donors through Ficoll-Paque gradient centrifugation. Cytokines were purchasedfrom Peprotech Inc., and all other tissue culture supple-ments from Life Technologies, and chemicals from SigmaAldrich (St. Louis, MO) unless otherwise specified. Com-pounds Ruxolitinib and sc144 were a generous gift fromthe laboratory of Dr. Karen McLean. Viral vectors werepurchased from Sigma Aldrich and packaged at the Uni-versity of Michigan Viral Vector core.

Derivation and polarization of macrophages from U937cell line, and PBMCsU937 cells were cultured in suspension in RPMI supple-mented with 10% heat-inactivated fetal calf serum (At-lanta Biologics) and 1x antibiotics/antimycotics. Cellswere harvested, and suspended at 2500cells/ml andtreated with 5 ng/ml phorbol myristate acetate (PMA).20 μl of this suspension was plated onto each well of ahanging drop array plate, to allow monocytes to differ-entiate into macrophages in suspension culture. ForPBMCs, cells were plated onto tissue culture dishes, andthe non-adherent cell fraction was discarded following24 h of attachment. PBMCs were then detached fromthe plate, and plated onto hanging drop arrays at 500cells/drop. At the end of 24 h, each well was left un-treated to derive M0 resting macrophages, or treatedwith 20 ng/ml recombinant human M-CSF and 20 ng/mlIL-4 to derive activated M2-like macrophages for thenext 48 h. For brevity, figure captions refer to IL-4/

Raghavan et al. Journal for ImmunoTherapy of Cancer (2019) 7:190 of 15



j.com/

J Imm


ownloaded from


MCSF activated macrophages as M2, to indicate theM2-like, alternatively activated phenotype. Macrophageswere harvested from hanging drops and assessed for dif-ferentiation and polarization using flow analysis, de-scribed below. Harvested macrophage aggregates werealso subjected to subsequent qPCR analysis, or used tomanufacture hetero-spheroids.

Isolation of ovarian CSCs from ovarian cancer cell linesOvarian CSCs were isolated from the serous ovarian can-cer cell lines OVCAR3 (used under passage 35) as de-scribed previously [13]. Briefly, cells were harvested andincubated with ALDEFLUOR reagent, and CD133 anti-body, and sorted using flow cytometry for cells positivefor elevated ALDH and CD133 positivity. AppropriateDEAB and isotype controls were used for both assays, todetermine gate settings as described previously. CSCswere freshly sorted and used to make hetero-spheroids <24 h after flow sorting.

Formation of mono- and hetero-spheroids from CSCs andmacrophagesSpheroids were generated on a hanging drop array platefrom CSCs and macrophages adapting protocols describedpreviously [13–15]. For mono-spheroids, 100 CSCs or theunsorted bulk of OVCAR3 cells were seeded per hangingdrop and allowed to form spheroids. Macrophages wereharvested from hanging drops following the differentiationprotocol described in Section 2.2. Macrophages and CSCswere combined and plated onto hanging drop arrays suchthat each drop contained 100 CSCs and 100 M0/M2 mac-rophages. Mono-spheroids contained 100 CSCs or bulk un-sorted OVCAR3 cells. Following 4–5 days in hanging droparray culture, spheroid formation was tracked using live cellmicroscopy and routinely fed to maintain a ~ 20 μl dropvolume. Spheroids were used for subsequent flow analysis,qPCR or lysed to obtain protein for immunoblots. In someinstances, hetero-spheroids were manufactured from CSCsstably expressing GFP, and fluorescent activated cell sortingwas utilized to separate the GFP+ CSC compartment fromhetero-spheroids for further analysis.

Flow cytometry analysisFlow cytometry analysis was performed following proto-cols established in our lab previously [13]. CD68-APC(Miltenyi Biotech, Germany) antibody was used, with itsassociated APC-isotype control, to identify CD68+ mac-rophages, in differentiated U937 or PBMC monocytes,and hetero-spheroids. For CD68 flow analysis, sampleswere fixed in methanol at − 20 °C, for 1 h, followed by aPBS wash to remove methanol, re-suspension in FACSBuffer (PBS + 2% FBS) before antibody incubation.Two kinds of flow cytometry based experiments were

carried out to characterize hetero-spheroids: i) macrophage

polarization was assessed by CD68+ and CD206+ using flowcytometry; ii) Stemness was assessed by ALDEFLUORassay for observation of elevated ALDH activity using flowcytometry, using protocols established previously [13].Briefly, hetero-spheroids were harvested in FACS buffer,and triturated to single-cell suspensions. Appropriateisotype controls were used for conjugated antibodies, to setgates to observe CD68 and CD206. For ALDH, a molar ex-cess of the DEAB inhibitor was used to determine positivegates per the manufacturer’s protocol, to identify elevatedALDH activity. ALDH activity was assessed following 48-htreatment with the JAK1/2 inhibitor, Ruxolitinib, or theGP130 inhibitor, sc144, or the human anti IL-6 antibody,Tocilizumab (Actemra, Genentech). Flow cytometry wasperformed on the Attune acoustic focusing flow cytometer(Applied Biosystems). Flow sorting was performed on theAstrios (Beckman Coulter).

Gene expression via qPCRRNA was extracted from harvested macrophages or hetero-spheroids using the RNeasy extraction kit (Qiagen). Ex-tracted RNA was assessed for concentration and purityusing a Nanodrop 2000 (Thermo Fisher Scientific) spectro-photometer. RNA was transcribed to cDNA using theHigh-fidelity cDNA Transcription kit (Life Technologies),and qPCR was carried out in the 96-well format using the7900HT platform (Applied Biosystems). For macrophagepolarization, CD163 and CD206 were assessed. IL-10 wasadditionally assessed in hetero-spheroids. Lastly, WNT li-gands were also assessed in hetero-spheroids. Gene expres-sion differences were quantified using the 2ΔΔCT method,using GAPDH as the housekeeping control, and reportedas fold change compared to a control sample. For macro-phages, controls were undifferentiated monocytes. Forhetero-spheroids, control samples were bulk OVCAR3spheroids. qPCR experiments were run in triplicates, with2–3 independent samples. A list of primers used in theqPCR experiments is provided in Additional file 1: TableS1.

Quantification of cytokines using ELISAFor ELISA assays, media was harvested from 50 spher-oids (macrophages, OVCAR3, CSC, CSC/M2 or CSC/sh-WNT5B M2). ELISA assays were performed on theDuoset ELISA system (R&D Biosystems, MinneapolisMN) following the manufacturer’s protocol, modified toinclude an overnight sample incubation. Cytokines ana-lyzed included IL-10 and IL-6. Standard curves weregenerated for each cytokine, and analyte concentrationwas assessed using a four parametric ELISA curve, todetermine the amount of IL-10 or IL-6 released. ELISAassays and data analysis were performed at the Immuno-logical Monitoring Core at the Rogel Cancer Center,University of Michigan.




j.com/

J Imm


ownloaded from


Assessment of chemoresistance in hetero-spheroidsFor chemoresistance, hetero-spheroids were treated withcarboplatin, to a final concentration of 300 μM, within20 μl drops, for 48 h. At the end of 48 h, the MTS re-agent (Abcam) was added to drops at a 1/10 dilution,and allowed to incubate at 37 °C for 2.5 h. At the end ofthe incubation period, absorbance was read on thehetero-spheroids at 590 nm, according to manufacturer’sprotocols. Untreated hetero-spheroids were used as con-trols, to normalize the absorbance to identify the effectof drug treatment on cellular viability. Results werequantified as normalized cell viability, based on un-treated controls. Experiments were repeated with 3–5biological replicates for statistical analysis.

Assessment of migration of hetero-spheroidsIn order to quantify invasiveness of hetero-spheroids, 8 μmtranswell inserts were placed in each well of a 24 wellplate. 10 CSC mono-spheroids or CSC/M2, CSC/scrambleM2, CSC/sh-WNT5B M2 hetero-spheroids were harvestedat Day 5 from hanging drop arrays, and placed on the topchamber of a transwell insert. The bottom chamber wasfilled with 400 μl of fresh medium, so only the bottom ofthe transwell insert was immersed in medium. Following 3days, the transwell insert was removed, and several imagesof the bottom of the 24 well were obtained using phasecontrast microscopy. Image J was used to quantify thenumber of cells in a field of view. At least four randomnon-overlapping fields of view were counted from each ex-periment, to find the number of cells that migrated throughthe transwell insert to the bottom of the well.

Immuno-blotting for β-cateninHetero-spheroids were harvested and lysed in 200 μl ofRadio-immunoprecipitation assay (RIPA) Buffer, soni-cated for 30s on ice with a probe sonicator. Extractedconcentration was measured using the BCA Assay Re-agent (Pierce) following manufacturer’s protocol for a96-well format. Subsequently, 50 μg of protein from eachsample was loaded onto 4–20% gradient polyacrylamidegels (Biorad), and separated electrophoretically, trans-ferred to a PVDF membrane. Transferred membraneswere blocked with 5% non-fat milk, and probed with β-catenin (R&D Biosystems) overnight at 4 °C, washedwith TBST buffer, and probed with an appropriate HRP-conjugated secondary antibody. β-Actin was used as aloading control to determine changes in β-catenin ex-pression among samples. ECL reagent (Pierce ProteinBiology) was used to visualize bands in a Biorad Chemi-Doc Touch instrument. Digital images acquired wereprocessed through NIH Image J, and band analysis toolswere used for densitometry. Band densities were normal-ized against the loading control β-Actin, to determinechanges.

Knock-down of WNT5B in macrophagesMission shRNA plasmids were obtained transformed intoE. coli from Sigma Aldrich, targeted to WNT5B(TRCN0000123194). Transformed E. coli were grown in LBmedium. Plasmid DNA was isolated using the PromegaDNA isolation kit following manufacturer’s protocols, and2.5 μg of DNA was transfected along with packaging pro-viral plasmids into HEK293-T cells. Lentivirus particleswere isolated at 1X concentration by the University ofMichigan Viral Vector Core. 1 × 105 cells were transducedwith 3 μg/ml polybrene and 0.5X lentivirus for 30min at800 g, at 32 °C in a centrifuge. Resulting pellets were re-suspended in fully supplemented growth medium for 72 h.At the end of 72 h, cells were harvested for qPCR analysis,or for macrophage differentiation and further experimenta-tion. Lentiviruses were packaged to express shRNA target-ing WNT5B (sh-WNT5B), or scrambled shRNA (sh-scramble). Sh-WNT5B or sh-scramble treated U937 mono-cytes were differentiated and activated into macrophagesfollowing protocols described in section 2.2.

Assessment of tumorigenicity in vivo of hetero-spheroidsCSC mono-spheroids and CSC/M2 and CSC/sh-WNT5BM2 hetero-spheroids were generated following protocolsoutlined in Section 2.4. CSCs were GFP tagged in thesespheroids, and after five days of hetero-spheroid culture,CSCs were isolated using the GFP label prior to sub-cutaneous injection into NSG mice. Each tumor receivedCSCs from 10 spheroids. Tumor initiation and monitoringprotocols were performed as described previously [13].When palpable tumors were observed, tocilizumab (10mg/kg, intra peritoneally) treatment began 3 times/week.Tumors were allowed to grow till the end-point wasreached for maximum tumor burden, and mice were eu-thanized. Tumors were dissected, and routine paraffinhistology and H&E staining was performed, to understandany changes in histology. RNA was isolated from tumorsfollowing protocols outlined in Section 2.6, and subject toqPCR for ALDH1A1, and several other Wnt ligands.

Data analysis and statisticsExperiments were carried out using 3–5 biological repli-cates for U937-derived macrophages, and OVCAR3-derived CSCs. GraphPad Prism 5.0 (www.graphpad.com)was utilized to perform all statistical analysis. When ap-propriate, one-way ANOVAs were used to test signifi-cant differences, and if differences were observed,indicated with symbols and a significance level.

ResultsMonocyte-derived macrophages can be differentiatedand activated in 3D hanging drop culturesThe monocytic cell line, U937, or healthy-donor derivedPBMCs were placed in hanging drop cultures. In the




j.com/

J Imm


ownloaded from

http://www.graphpad.com


presence of PMA, U937 and PBMC monocytes differ-entiated into macrophages, and over the course of 72h, they were organized as a compacted mass of cellswithin hanging drop cultures (Fig. 1a). Monocyteswith no PMA stimulation, in contrast, were extremelyloosely aggregated and did not form compact spher-oids. Differentiated macrophages were indicated asM0 macrophages, implying differentiation with nocytokine stimulation. 74.6 ± 8.2% of U937 monocytesdifferentiated into M0 macrophages expressed the panmacrophage marker CD68. Similarly, PBMCs also dif-ferentiated into macrophages upon PMA stimulation,with 84.3 ± 8.5% CD68 expression (Fig. 1b). Mono-cytes were also differentiated and polarized into an al-ternatively activated phenotype (M2; Fig. 1a). Thesemacrophages were derived with PMA stimulation inthe presence of IL-4 and M-CSF. Alternately activatedM2-like macrophages, either from U937 or PBMCshad an associated increase in gene expression ofCD163 and CD206 (Fig. 1c). Immuno-suppressivecytokine IL-10 and tumor-promoting cytokine IL-6were elevated in M2-polarized macrophages fromboth U937 and PBMC compared to M0 macrophages,or undifferentiated monocytes, indicating a shift inphenotype between M0 and M2-like macrophages de-rived using this culture system (Fig. 1d). In summary,macrophages could be derived from U937 or PBMCmonocytes using the 3D hanging drop culture system,and could be further activated into a M2-phenotype.3D differentiation and activation was similar in termsof gene expression to conventionally activated macro-phages in 2D culture systems, including elevated argi-nase enzyme activity (Additional file 1: Figure S1).

Hetero-spheroids can be derived from ovarian cancerstem cells and macrophages using hanging drop culturesHetero-spheroids combining ovarian CSCs derivedfrom the OVCAR3 cell line and monocyte-derived mac-rophages (U937 or PBMC) in a starting ratio of 1:1were generated using hanging drop arrays. Phase con-trast images of hetero-spheroids at Day 5 indicate com-pact spheroid formation with tight defined boundaries(Fig. 1e). No significant differences in size or prolifera-tive index were observed between CSC mono-spheroids, and CSC/M0 or CSC/M2 hetero-spheroidsAdditional file 1: Figure S2). Hetero-spheroids retainedbetween 19.2 ± 1.5%-26.3 ± 1.8% expression of CD68+

macrophages at Day 5 (Fig. 1f ). Gating strategy for flowanalysis is presented in Additional file 1: Figure S3.CSC/M0 or CSC/M2 hetero-spheroids therebymaintained robust macrophage populations within thespheroids, while maintaining compact spheroid archi-tecture for 5 days.

Ovarian cancer stem cells drive CD206 expression in M0monocyte-derived macrophages within hetero-spheroidsthrough IL10 and Wnt-signalingIn order to understand if there are differences in bulkovarian cancer cells (OVCAR3), and ovarian cancerstem cells (ALDH+ CD133+ CSC) in their ability todrive an immuno-suppressive macrophage phenotype,we generated hetero-spheroids from OVCAR3/M0and CSC/M0. OVCAR3 or CSC mono-spheroids dem-onstrate minimal expression of CD206 (0.9–1%,Fig. 2a, Additional file 1: Figure S4). When comparingdifferences in CD206 expression between OVCAR3/M0 or CSC/M0 hetero-spheroids, a significant (**p <0.001, one-way ANOVA) 20% increase in CD206 wasobserved with CSC co-culture (Fig. 2a), indicatingthat ovarian CSCs drive an immuno-suppressivephenotype in macrophages compared to bulk ovariancancer cells. CSC/M2 hetero-spheroids also expressedanother alternately activated M2 macrophage marker,CD163 (Additional file 1: Figure S5). We sought toexplore whether there are differences in the immuno-suppressive cytokine IL-10 between OVCAR3 andCSC spheroids. We found that IL-10 gene expressionwas significantly elevated (2-fold) in CSC mono-spheroids compared to OVCAR3, which was evenmore pronounced in CSC/M0 spheroids compared toOVCAR3/M0 (Fig. 2b). In hetero-spheroids derivedfrom M2 macrophages and CSCs, CD206 expressionwas maintained through co-culture (Fig. 2c). Highersecreted IL-10 levels were observed in CSC/M2hetero-spheroids (Fig. 2d).We then explored the possibility of a factor other than

IL-10 driving CD206 expression in macrophages. Giventhe importance of the WNT signaling cascade in bothOvarian CSC maintenance as well as macrophagepolarization and activation, we inhibited all WNT secre-tion from Ovarian CSCs using an inhibitor, IWP-2.When hetero-spheroids were generated from IWP-2treated CSCs and M0 or M2 macrophages, CD206 ex-pression significantly decreased (44–62%) in hetero-spheroids (*p < 0.05, one-way ANOVA; Fig. 2e), with noassociated change in CD68 expression. Concomitantly,we found the gene expression of several Wnt ligands ele-vated close to 2-fold in CSC compared to bulk OVCAR3spheroids (Fig. 2f ). Furthermore, we found an associated30% increase in beta-catenin protein expression, indicat-ing an increased canonical Wnt signaling axis in CSC/M2 spheroids (Fig. 2g).In short, in CSC-macrophage interactions within

hetero-spheroids, we found not only elevated levels ofthe immuno-suppressive cytokine IL-10, but also ofseveral WNT ligands. We also noted a dependency ofM2 macrophage activation on CSC-derived WNTligands.




j.com/

J Imm


ownloaded from


Alternatively activated M2 macrophages increase ovarianCSC populations within hetero-spheroids through IL-6signaling, and are more chemoresistant and invasiveWe then investigated the maintenance of stemness withinhetero-spheroids, since we hypothesized that alternativemacrophage activation may result in a pro-tumoral reci-procity within hetero-spheroids. Flow analysis for elevatedALDH indicated that M2 hetero-spheroids significantlyand robustly (**p < 0.001, one-way ANOVA) increased themaintenance of ALDH+ populations within hetero-spheroids, compared to CSC mono-spheroids (Fig. 3a).Macrophages themselves do not significantly express

elevated ALDH in this assay (Additional file 1: Figure S6).A ~ 2 fold increase in ALDH+ populations was observedwith co-culture with either U937M2 or PBMC M2 hetero-spheroids, indicating that M2 activation resulted in im-proved ovarian CSC maintenance. Concomitant with thisincrease in ALDH, we also observed increased secretion ofthe IL-6 pro-tumoral cytokine (Fig. 3b). Consequently, in-hibition of IL-6 signaling with two small molecule inhibi-tors, Ruxolitinib or SC144 during the formation of CSC/M2 hetero-spheroids significantly (**p < 0.001, one-wayANOVA) reduced the enrichment of ALDH+ cells (Fig. 3c).Our results indicate that the IL-6 signaling axis initiated by

Fig. 1 Hetero-spheroids derived from monocyte-derived macrophages and ovarian cancer stem cells. a Monocytes from the U937 cell line orperipheral blood monocytes (PBMC) were plated into hanging drop arrays, and differentiated to M0 macrophages with phorbol ester treatment,or activated with IL-4 and MCSF treatment. Differentiated and activated macrophages formed compact spheroid-like aggregates. b DifferentiatedM0 and M2 macrophages all expressed pan macrophage marker, CD68 indicating 75–80% differentiation efficiency from monocytes tomacrophages. CD68 expression was evaluated using flow cytometry analysis, with representative plots. c Polarization was assessed using qPCRanalysis for two M2 genes, CD163 and CD206. Both U937 and PBMC macrophages expressed significantly higher levels of CD163 and CD206genes, compared to untreated undifferentiated monocytes. d M2 differentiated macrophages secreted higher amounts of the immuno-suppressive cytokine IL-10, and the pro-tumoral cytokine IL-6. e CSCs were derived from the OVCAR3 cell line based on ALDH+ CD133+

expression. Hetero-spheroids were generated using differentiated U937 or PBMC M0 macrophages and CSCs or activated U937 or PBMC M2macrophages and CSCs. Representative phase contrast images of hetero-spheroids seen at Day 5 following formation indicate compactspheroids, similar in size to CSC mono-spheroids generated from the same number of CSCs/spheroid. f Hetero-spheroids retain ~ 20% CD68expression, indicating that at day 5, CD68+ macrophages constitute 20% of the population of cells. Scale bar = 200 μm




j.com/

J Imm


ownloaded from


M2 macrophage co-culture significantly increases mainten-ance of ALDH+ CSCs within hetero-spheroids.As a consequence of ALDH+ enrichment, CSC/M2

hetero-spheroids were markedly (**p < 0.001, p < 0.05,one-way ANOVA) chemoresistant to carboplatin treat-ment (Fig. 3d). Phase contrast images of drug-treatedspheroids demonstrated the loss of integrity of the tightspheroid borders in CSC mono-spheroids, whereas theloss of compactness and integrity was less visually obviousin CSC/M2 hetero-spheroids (Fig. 3d). Concomitant withincreased drug resistance, CSC/M2 hetero-spheroids were

also significantly (*p < 0.05, one-way ANOVA) more inva-sive (2–2.6 fold) in transwell assays compared to CSCmono-spheroids (Fig. 3e).Our experiments indicated that culturing CSCs within

CSC/M2 hetero-spheroids resulted in increased main-tenance of ALDH+ CSC populations, which translated tofunctional increases in chemoresistance, and invasive-ness of CSC/M2 hetero-spheroids. Further, blocking in-creased IL-6 related signaling in CSC/M2 hetero-spheroids with small molecule inhibitors reduced themaintenance of CSCs within hetero-spheroids.

Fig. 2 Hetero-spheroids drive CD206 polarization in monocyte-derived macrophages. a Hetero-spheroids were generated from bulk unsortedOVCAR3 cells and M0 macrophages (OVCAR3/U937M0), and CSC/U937M0. Flow analysis for the macrophage polarization marker CD206indicated that while OVCAR3 and CSCs by themselves do not express CD206, CSCs drive CD206 expression in previously CD206− M0macrophages within hetero-spheroids at significantly (**p < 0.001, one-way ANOVA) higher levels than OVCAR3. Representative flow analysis plotsare shown. b IL10 gene expression was evaluated using qPCR in mono- and hetero-spheroids, indicating that CSCs had a 2-fold increased geneexpression of IL10, likely driving CD206 expression in M0 macrophages to higher extents than OVCAR3. c CD206 expression was maintained atelevated levels (18.5–29.24%) in all CSC hetero-spheroid conditions, indicating that the activated M2 phenotype was maintained in macrophageswithin hetero-spheroids. d Levels of secreted IL-10 were mildly elevated in CSC/M2 hetero-spheroids, but were also similar across all CSC/macrophage hetero-spheroids, indicating the presence of the immuno-suppressive cytokine. e When CSCs were pre-treated with the Wntsecretion inhibitor, IWP-2, significantly lower (*p < 0.05,one-way ANOVA) CD206 expression was observed in CSC/M2 hetero-spheroids, implyingthe importance of CSC-derived Wnt ligands in the maintenance of M2 activation in macrophages. f qPCR analysis revealed that gene expressionof several Wnt ligands were elevated in CSC compared to bulk OVCAR3 spheroids (dotted line = no change); (g) Densitometric measurements ofimmunoblots for β-catenin compared to the loading control β-actin indicated that CSC/M2 spheroids had a 26% increase in β-catenin expression,compared to CSC mono-spheroids, indicating higher Wnt/β-catenin activity in CSC/M2 hetero-spheroids




j.com/

J Imm


ownloaded from


Increased WNT signaling in macrophages is responsiblefor the pro-tumoral and immuno-suppressive phenotypeobserved in M2-hetero-spheroidsWe hypothesized that similar to CSCs, M2-like macrophageactivation may also result in increased WNT signaling. Weblocked pan-WNT ligand secretion in M2 macrophages bytreatment with IWP-2 following polarization. CSC/IWP2M2 hetero-spheroids assumed different aggregation morph-ology and were less compact (Fig. 4a). Furthermore, ALDHexpression was significantly lower (**p < 0.001, one-wayANOVA) in CSC/IWP2 M2 hetero-spheroids compared toCSC/M2, with minimal reduction in CD206 expression(Fig. 4b), indicating that macrophage-derived WNT signal-ing may at least be partially responsible for the phenotypes

observed in ovarian CSCs. We investigated the expressionof WNT ligands in macrophages, and found WNT5B over-expressed ~ 32 fold in M2 macrophages compared tomonocytes using qPCR. We used an shRNA plasmid target-ing the WNT5B ligand (sh-WNT5B) and a non-targetedplasmid (sh-Scramble), and used lentiviral methods to si-lence WNT5B expression in monocytes, with a knockdownefficiency of 76% (Fig. 4c). Macrophages derived from sh-WNT5B were termed sh-WNT5B M2. Despite > 75%knock-down efficiency of WNT5B in monocytes, AlternateM2-like activation still resulted in an increase in gene ex-pression of WNT5B. However, we found that monocytestreated with sh-WNT5B showed a 52% reduction inWNT5B expression compared to control or scramble M2

Fig. 3 M2-polarized macrophages increase ALDH+ in CSC/M2 hetero-spheroids. a Elevated ALDH activity was assessed using the ALDEFLUORflow cytometric analysis assay, with representative flow plots demonstrated. CSC/M2 hetero-spheroids, whether U937 or PBMC, have significantly(**p < 0.001, one-way ANOVA) elevated ALDH+ expression (1.8–2.4 fold). b Secreted pro-tumoral cytokine IL6 is elevated in hetero-spheroidscompared to CSC mono-spheroids. c We suppressed signaling through the IL6/STAT3 axis using two inhibitors Ruxolitinib and sc144 in hetero-spheroids, and observed that this suppression also significantly (**p < 0.001, one-way ANOVA) decreased the ALDH+ enrichment within CSC/M2hetero-spheroids. d Concomitant with the ALDH expression within hetero-spheroids, sensitivity to carboplatin was evaluated using the MTS assay.Normalized absorbance values indicated that CSC/M2 hetero-spheroids were significantly (*p < 0.05, one-way ANOVA) more resistant tocarboplatin treatment compared to CSC mono-spheroids made with the same number of CSCs. e Representative phase contrast images indicatethe loss of the tight spheroid boundaries in the CSC mono-spheroids in response to carboplatin treatment, indicative of cell death. The extent ofthe loss of boundary integrity is much lower in the CSC/M2 hetero-spheroids. Scale bar = 100 μm. f Quantification of the number of migratedcells in the lower chamber of a transwell system indicated that CSC/M2 hetero-spheroids were significantly (*p < 0.05, one-way ANOVA) moremigratory compared to CSC mono-spheroids




j.com/

J Imm


ownloaded from


macrophages (Fig. 4d). WNT5B knockdown did not alterthe CD68+ marker expression (Fig. 4e). Treatment with IL-4/M-CSF to induce macrophage polarization resulted in in-creased gene expression for CD163 and CD206, indicatingthe development of an alternate M2-like phenotype (Fig.4f). Therefore, knockdown of WNT5B using sh-WNT5Bdid not significantly alter the differentiation of monocytes tomacrophages, or the development of an M2 activatedphenotype. However, a significant loss in WNT5B gene ex-pression was observed.

WNT5B is involved in M2 macrophage-CSC interactions,leading to chemoresistance, invasiveness, and increasedstemnessHetero-spheroids were generated with CSCs and Scram-ble M2 or sh-WNT5B M2. No significant differences inCD206 expression was observed in these hetero-spheroids, indicating that in co-culture, macrophageknockdown of WNT5B was dispensable for the expres-sion of CD206 (Fig. 5a). However, macrophage knock-down of WNT5B resulted in a significant (*p < 0.05,

Fig. 4 Inhibition of macrophage Wnt-secretion reduces ALDH enrichment in hetero-spheroids. a Representative phase contrast image of a hetero-spheroid generated from CSCs and IWP-2 treated U937M2 macrophages, indicates the formation of aggregated spheroids. Scale bar = 200 μm. b Flowanalysis revealed that hetero-spheroids with IWP-2 treated M2 macrophages had a significantly diminished ALDH+ enrichment (**p < 0.001, one-wayANOVA) compared to control untreated CSC/M2 hetero-spheroids. However, no change in CD206 expression was observed in hetero-spheroids. cTransduction efficiency of monocytes with shRNA directed against WNT5B indicated a > 75% efficiency in knockdown of WNT5B gene expression inthe shWnt5b targeted construct, and minimal changes in WNT5B gene expression in the scramble non-targeted lentiviral construct. d We utilizedshWnt5b monocytes to differentiate and polarize M2 macrophages, and found that there was a 52% reduction in WNT5B gene expression in shWnt5bM2 macrophages, compared to scramble or control untreated M2 macrophages, indicating the knockdown of the Wnt5b gene. e No changes wereobserved in CD68 expression in monocytes generated from scramble or shWNT5B treated monocytes, demonstrated by flow analysis andrepresentative plots. f Similarly, qPCR analysis of M2 gene expression markers CD163 and CD206 indicated increases in both genes, indicative ofactivation into the M2 program even in shWnt5b macrophages




j.com/

J Imm


ownloaded from


one-way ANOVA) decrease in ALDH+ compartment byover 5-fold, indicating that macrophage-derived WNT5Bwas crucial for the enrichment of CSC characteristicswithin hetero-spheroids (Fig. 5b). This reduction in theALDH+ enrichment was also associated with an in-creased sensitivity to carboplatin (Fig. 5c), and a de-creased invasive potential (Fig. 5d). Upon examination ofcytokine profiles, we observed no significant differencesin IL-10 (Fig. 5e). However, the pro-tumoral CSC-promoting cytokine IL-6 was completely abrogated inthe CSC/sh-WNT5B M2 hetero-spheroids, indicatingthat WNT5B was likely also driving CSC phenotypesthrough IL-6 secretion. We explored the possibility ofrestoring the CSC enrichment phenotype by the additionof exogenous IL-6. Upon exogenous addition of IL-6,levels of ALDH+ cells increased within CSC/sh-WNT5BM2 hetero-spheroids, indicating that IL-6 is a key ef-fector downregulated upon WNT5B knockdown, butmay likely not be the only mediating factor in promotingCSC maintenance via WNT5B (Fig. 5f ). Concomitantwith the loss of IL-6 in CSC/sh-WNT5B M2 hetero-spheroids, we also observed reduced phosphorylatedSTAT3 (42.4 ± 5.5%) in immunoblots compared to CSC/M2 hetero-Additional file 1: Figure S7). Lastly, we ex-plored the possibility of WNT-driven WNT signaling inCSCs, in response to macrophage WNT5B. We ob-served through gene expression analysis that severalWNT ligands were significantly over-expressed in CSCsco-cultured with M2 macrophages, and there was a lossin WNT ligand expression upon co-culture with sh-WNT5B M2 macrophages (Fig. 5g). Consequently, therewas also an associated ~ 50% loss of β-catenin proteinexpression in CSC/sh-WNT5B M2 hetero-spheroids, in-dicating a lower paracrine WNT activation in CSCs inco-culture with macrophages where WNT5B wasknocked down (Fig. 5h).Overall, we observed that although knockdown of

macrophage WNT5B did not alter macrophage M2phenotype, macrophage-derived WNT5B proved to be adriver of WNT-activated WNT signaling and IL-6 secre-tion, responsible partly for the increased CSC pheno-types observed in CSC/M2 hetero-spheroids.

CSCs conditioned in M2 hetero-spheroid culture are moretumorigenic in vivo, and shWNT5B macrophageconditioning inhibits tumorigenicityThree tested conditions: CSC (mono-spheroids), CSC (M2;conditioned in hetero-spheroids) and CSC(sh-WNT5B M2;conditioned in hetero-spheroids) generated tumors in NSGmice (Fig. 6a). The rate of tumor initiation and kinetics ofgrowth was significantly different between all three condi-tions, with CSC(M2) tumors initiating at the fastest rate of11.97mm3/day. CSC(shWNT5b M2) tumors were signifi-cantly slower to initiate at 4.74mm3/day, compared to CSC

tumors at 5.89mm3/day. CSC(M2) tumors reached themaximum tumor burden window earlier, compared toCSC and CSC(shWNT5B M2) tumors (Fig. 6a). Histologicexamination of the xenografts reveals that the lesions arecomposed of solid sheets of tumor cells with epithelioidmorphology. In the CSC group, dense sheets of polygonaltumor cells with abundant cytoplasm and conspicuousnuclei are noted, with little intercellular space (Fig. 6b). Inthe CSC(M2) group, sheets of large epithelioid tumor cellswith abundant eosinophilic cytoplasm constitute the en-tire tumor with little intercellular space and stroma. In theCSC(shWNT5B M2) group, the tumor cells are muchsmaller and loosely arranged in cords and strands. Occa-sional apoptotic cells are seen with a low mitotic count.Gene expression levels were assessed in xenografted tu-mors from the three groups, to understand if macrophageconditioning was maintained in vivo (Fig. 6c). Our data in-dicated that an elevated ALDH1A1 expression wasretained significantly in CSC(M2) tumors (**, p < 0.001,two-way ANOVA), compared to CSC(shWNT5B M2) tu-mors (ns; not significant). Similarly, elevated Wnt ligandsobserved in vitro upon CSC/macrophage co-culture inhetero-spheroids (Fig. 5g) was re-evaluated in xenografts.We observed the maintenance of a significant elevation inWnt2, Wnt3A, and Wnt6 (*p < 0.05, **p < 0.001, two-wayANOVA) in CSC(M2) tumors compared to CSC tumors.CSC(shWNT5B M2) tumors were not significantly differ-ent in Wnt ligand elevation compared to CSC tumors.Furthermore, we tested the effect of the human IL-6 in-hibitor Tocilizumab on CSC and CSC(M2) spheroids (Fig.6d). CSC spheroids were significantly responsive to Toci-lizumab treatment, evident from the divergent treatmentcurve from the control untreated group (purple curves,Fig. 6d). CSC(M2) tumors were however significantly re-sistant to Tocilizumab treatment, indicating the elevatedIL6 conditioning activity, in line with elevated ALDH1A1gene expression and elevated Wnt ligand expression.Change in tumor burden with Tocilizumab treatment inCSC spheroids ranged between 67.7–70.7%, while inCSC(M2) tumors, the reduction was merely 23.8–24% ofcontrol tumor volumes.

DiscussionMacrophages play an important role in tissue homeostasis,with tumor-associated macrophages promoting cancer pro-gression, metastasis, angiogenesis and tumorigenicity.Tumor-associated macrophages are a major fraction ofcells within the epithelial ovarian cancer ascites microenvir-onment [23, 24], and are responsible for recurrence andmetastasis of ovarian cancer, including promotion of resist-ance and the preservation of a more de-differentiated, i.e.stem-like cancer cell phenotype [7, 25]. Given the enrichedpresence of CSCs and the abundant presence of macro-phages within the malignant ascites [1, 2, 11], we opted to




j.com/

J Imm


ownloaded from


use a hetero-spheroid model to bring both these cells inclose association within one 3D structure in vitro. The pri-mary advantage of this model is that it facilitates cell-cellinteractions in a 3D microenvironment [13–15]. Therefore,the hanging drop spheroid model is uniquely positioned tounderstand how ovarian CSCs and macrophages interactwith each other (simulating their presence within the asci-tes), and specifically, test our hypotheses: i) CSCs activatemacrophages differently comparing to the bulk ovarian

cancer cells; and ii) the activated macrophages promotefunctional CSC phenotypes.In our experiments, we observed similar levels of dif-

ferentiation and subsequent activation of macrophagesby using the 3D platform, compared to monocytes dif-ferentiated and activated traditionally in 2D tissue cul-ture plates by us (Additional file 1: Figure S1) and others[26–28]. Interestingly, along with clear hallmarks of al-ternate macrophage activation (IL-10 secretion, CD163

Fig. 5 Macrophage knockdown of Wnt5b suppresses ALDH enrichment, with no change in CD206, and increases chemosensitivity of hetero-spheroids. a Flow analysis of CD206 indicated no significant differences in CSC/M2 and CSC/Scramble M2 and CSC/shWnt5b M2 hetero-spheroids, indicating that CSCs were still capable of maintaining a robust immunosuppressive program in hetero-spheroids, regardless of WNT5B.b ALDH elevation however was significantly (*p < 0.05, one-way ANOVA) lowered compared to CSC/U937 M2 hetero-spheroids, and notsignificantly different (ns) compared to CSC mono-spheroids, indicating that knocking down Wnt5b expression in macrophages reduced ALDHenrichment in hetero-spheroids. c Concomitant with the decrease in ALDH, sensitivity to carboplatin also significantly (**p < 0.001, one-wayANOVA) improved, responding to treatment similar to CSC mono-spheroids. The red dashed line indicates the sensitivity levels of CSC mono-spheroids. d CSC/shWnt5b M2 hetero-spheroids were also significantly (***p < 0.0001, t-test) less invasive than CSC/Scramble M2 hetero-spheroids. e Upon ELISA analysis of secreted IL10 and IL6, we found no changes in IL-10 levels, while we found a stark decrease in secreted pro-tumoral IL-6 cytokine in CSC/shWnt5b M2 hetero-spheroids. f Exogenous addition of IL-6 (25 ng/ml) to CSC/shWnt5b M2 hetero-spheroidspartially significantly (*p < 0.05, one-way ANOVA) increases ALDH+ enrichment, but does not restore it to levels of original enrichment with CSC/Scramble M2 or CSC/Ctrl M2 hetero-spheroids. g In order to identify if there was paracrine macrophage Wnt-driven Wnt signaling in CSCs, weseparated the CSCs (using a GFP label) from hetero-spheroids, and probed for gene expression of several Wnt ligands we saw elevated in CSCscompared to bulk OVCAR3. We observed that with M2 macrophage co-culture, several Wnt ligands were upregulated (Wnt2 significantlyelevated, ***p < 0.001, two-way ANOVA). However, this elevated gene expression of Wnt ligands was lost in CSCs in hetero-spheroid culture withshWnt5b M2, indicating that Wnt5b was partly responsible for potentiating Wnt signaling within the CSC compartment. h This loss in Wnt ligandexpression also translated to decreased β-catenin expression in CSC/shWnt5b M2 hetero-spheroids compared to CSC/U937M2 hetero-spheroids,quantified by densitometry of immunoblots (representative blot shown)




j.com/

J Imm


ownloaded from


and CD206 gene expression), we also observe an ele-vated macrophage secretion of IL-6, which is not typic-ally associated with M2 macrophages, but is observed inbipolar tumor associated macrophage populations in themalignant ascites of ovarian cancer [11, 29].In CSC/macrophage hetero-spheroids, we find that

CD68+ macrophages make up ~ 20% of the population atday 5, even though they start at 50% initially, likely due toCSC proliferation taking over the spheroid. Our resultssuggest that CSCs have an intrinsically higher immuno-suppressive program, driving CD206 expression in M0macrophages to a larger extent than bulk unsorted ovariancancer cells, evident also by the increased IL10 gene ex-pression in CSCs. This observation is in line with resultsobserved in literature where ovarian cancer cells activatemacrophages to alternative M2-like pro-tumoral phenotype

[9, 30]. Indeed, macrophages exposed to IL-10 are knownto polarize into M2 phenotypes [31, 32] under sustainedIL-10 exposure, and CSCs from ovarian and other cancerscan induce pro-tumoral macrophage phenotypes throughother pathways including NF-kB [33, 34].Consequently, we determined that hetero-spheroids

with alternatively activated M2-like macrophages in-creased maintenance of the ALDH+ CSC populations.This is in line with observations in ovarian, breast andhepatocellular carcinomas, where tumor-associated mac-rophages induce stemness and increase CSCs [25, 35,36]. In our experiments, ALDH+ increase could also becorrelated with increased levels of the cytokine IL-6 inM2 hetero-spheroids. The involvement of M2-derivedIL-6 is further strengthened by the fall in ALDH+ en-richment with inhibitors of the IL-6/STAT3 signaling

Fig. 6 Changes in tumorigenicity of hetero-spheroids in vivo in NSG mice. Tumors were generated in NSG mice from CSC mono-spheroids, orCSCs isolated from hetero-spheroids in CSC/M2 or CSC/shWNT5b M2. a Tumor initiation and growth kinetics are shown for all three groups,where CSC(M2) tumors (red curve) demonstrates elevated tumorigenicity, with faster tumor initiation and tumor burden development (grayshaded area indicates the tumor burden window). Similarly, the blue trace indicates the tumor growth in CSC/shWNT5b M2 tumors, which issignificantly (*p < 0.05, two-way ANOVA) lower and slower compared to CSC and CSC(M2) groups at the time points indicated. b Histologicalexamination demonstrates that all groups establish similar structures of tumors sub-cutaneously, with CSC tumors and CSC (M2) tumorsestablishing dense epitheliod cells, with CSC (shWNT5b M2) tumors more loosely arranged. Scale bar = 100 μm. c Gene expression analysis ofxenografted tumors indicated that an elevated ALDH1A1 expression was maintained in CSC(M2) tumors (**p < 0.001, two-way ANOVA). M2-likemacrophage conditioning also helped CSCs retain elevated expression of WNT2, WNT3A and WNT6 (*p < 0.05, **p < 0.001, two-way ANOVA)compared to CSCs in mono-spheroids. All other genes and conditions were not significantly different (ns) compared to CSCs, specifically CSC/shWnt5B M2 tumors. d The human IL-6 inhibitor Tocilizumab was used intra-peritoneally on CSC and CSC(M2) tumors, once palpable tumorswere observed (Day 25), indicated by the arrow. Tocilizumab treatment resulted in a significant decrease in tumor volume (**p < 0.001, two-wayANOVA) compared to control untreated tumors in CSC mono-spheroid tumors (compare purple trace to black trace; ~ 70% reduction in tumorburden). However, in CSC/M2 tumors, tocilizumab treatment did not significantly alter the growth kinetics of treated vs. untreated tumors (~ 24%reduction in tumor burden), indicating the detainment of a chemoresistant phenotype upon M2-like macrophage co-culture and conditioning.Furthermore, the treated CSC Tocilizumab condition also resembles the control CSC (shWNT5B M2) tumors (blue trace in A), indicating that partialloss or inhibition of IL-6 based CSC enrichment may be at play




j.com/

J Imm


ownloaded from


axis (Ruxolitinib, SC144, and Tocilizumab). The increasein ALDH+ enrichment and the emergence of CD206+

macrophages was also observed in CSC/M0-M2 hetero-spheroids generated from an additional high-grade serouscell line, Kuramochi (Additional file 1: Figure S8) andtumor cells and CSCs from a high-grade serous primaryovarian carcinoma sample Additional file 1: Figure S9).The WNT/β-catenin pathway is heavily implicated in

the maintenance of CSCs in ovarian cancer, so much so,that it is an attractive therapeutic target [37–39]. Hence,the elevation of several WNT ligands in CSCs comparedto bulk OVCAR3 cancer cells is unsurprising, as we ob-serve in our results. It is however likely, that some of theCSC-derived WNT ligands could potentially drive M2-like alternate macrophage activation, as evidenced in otherpathologies where the involvement of WNT3A, WNT6and other WNT ligands drive alternative macrophage ac-tivation [20, 40].In our experiments, we found that macrophage-

specific inhibition of WNT secretion resulted in a sig-nificant reduction in ALDH+ enrichment in hetero-spheroids, implying that WNT signaling was likely in-volved bi-directionally in CSC/macrophage interactions.The lone WNT ligand we found significantly upregu-lated in M2-like macrophages (compared to monocytesor M0 macrophages) was WNT5B, in line with tran-scriptomic and gene expression analyses performed onhuman M2 macrophages by other groups [41, 42]. Inter-estingly, macrophage WNT5B knockdown resulted inthe loss of ALDH+ enrichment in the CSC compartmentof hetero-spheroids, with an associated loss of IL-6 se-cretion. In other pathogenic states, activation of WNT5Bis associated with IL-6 secretion [43, 44], and hence, it isnot surprising that we observe the loss in IL-6 upon

WNT5B knockdown. Exogenous addition of IL-6 par-tially restores elevated ALDH levels in CSC/sh-WNT5BM2 hetero-spheroids, indicating that IL-6 is at least partlyresponsible for the observed phenotype. Additionally, wealso observed that several WNT ligands were upregulatedin the CSC compartment of hetero-spheroids cultured inCSC/M2 compared to CSC mono-spheroids, and the up-regulation dropped in CSCs cultured in CSC/sh-Wnt5bM2 hetero-spheroids. Tumor stroma-derived WNTligands like WNT3 and WNT5B are critical factors thatinstigate invasive behavior, and induction of an EMTphenotype in tumor epithelial cells [45]. In fact, WNT5Bassociated exosomes promoted cancer cell migration andproliferation in a paracrine manner [46].This points to enhanced Wnt signaling within the

CSC compartment upon co-culture with alternativelyactivated M2-like macrophages, likely mediated by se-creted Wnt (specifically also WNT5B) arising from themacrophage compartment, in line with several observa-tions where tumor associated/M2 macrophages increaseWnt signaling in epithelial cells [22, 47–49].The tumor-intrinsic WNT-β-catenin signaling is

shown to promote a T-cell exclusion phenotype [18].Utilizing a robust immune deconvolution algorithm, weanalyzed high-grade serous carcinomas and found a con-sistent phenotype, suggesting that the tumor-intrinsicWNT-β-catenin activation may promote immune sup-pression across cancer types (Additional file 1: FigureS10). However, the cellular mechanism underpinningthis link remained less characterized. This study employsa high-fidelity 3D culture model to reveal the reverselink between the expression levels of WNT5B in alterna-tively activated macrophages and cancer cells, possiblyleading to T-cell exclusion. However, it is important to

Fig. 7 Ovarian cancer stem cells and macrophages reciprocally interact through the Wnt pathway to promote a pro-tumoral microenvironment.Our data suggests that in hetero-spheroids, CSCs drive a CD206+ expressing macrophage phenotype, suggestive of pro-tumoral M2 activationthrough secretion of the immuno-suppressive cytokine IL-10, and through WNT ligands. We also observe that macrophage-derived WNT ligands,specifically WNT5B, and to some extent the pro-tumoral cytokine IL-6, drive an enrichment in ALDH+ cells within hetero-spheroids. Knockdown ofmacrophage WNT5B expression, or inhibiting downstream activation of IL-6 using Ruxolitinib or sc144, result in a significant loss of ALDH+

populations within hetero-spheroids. Importantly, we find that the Wnt pathway is involved bi-directionally in CSC-macrophage interactions, andcan potentially be targeted to reduce stemness, chemoresistance, invasiveness and immunosuppression in ovarian cancer




j.com/

J Imm


ownloaded from


note that the WNT signaling pathway has pleotropic ef-fects on a broad range of cell types including severaltypes of immune cells [50, 51]. Recent studies show thatWnt pathway is also crucial for the development of T-cells [52]. The Wnt-β-catenin pathway maintains themultipotency of memory CD8+ T-cells with anti-tumorproperties [53]. Thus, dosing, treatment schedule, andrelatively more specific tumor-targeted delivery ap-proaches are likely important to unleash the potential ofthe immune-priming properties of WNT-β-catenininhibitors.In conclusion, we demonstrate a robust hetero-

spheroid culture system, where ovarian CSCs and acti-vated macrophages can be brought in close associationin a 3D in vitro microenvironment, simulating theirnon-adherent presence within malignant ascites. Wedemonstrate the presence of macrophages within spher-oids and their pro-tumoral activation by CSCs. Recipro-cally, pro-tumoral activated macrophages also promote achemoresistant and invasive phenotype of the CSC com-partment and its enrichment within hetero-spheroids,potentially contributing to highly malignant and meta-static disease (Fig. 7). Lastly, we were able to understandthe reciprocal involvement of the WNT pathway, withactivation of paracrine WNT signaling by macrophagesin ovarian CSCs, with wide-implications for new thera-peutic targets to specifically eradicate the immuno-modulation of macrophages by CSCs that contribute torecurrent disease.

Additional file

Additional file 1: Figure S1. U937 monocytes differentiated in 3Dhanging drop arrays are equivalent to U937 monocytes differentiated in2D. Figure S2. No change in proliferation in CSC compartments ofhetero-spheroids. Figure S3. Gating strategy for Flow cytometry. FigureS4. Cancer cells do not significantly express the macrophage marker,CD206. Figure S5 CD163 expression is elevated in CSC/U937 M2 hetero-spheroids. Figure S6. Macrophages do not significantly express elevatedALDH. Figure S7. phospho-STAT3 is significantly reduced in CSC/shWNT5B-M2 hetero-spheroids compared to CSC/M2 hetero-spheroids.Figure S8. Kuramochi-CSC also drive elevated CD206 expression in mac-rophages, and polarized macrophages enrich ALDH+ cells in KuramochiCSC and resistance to carboplatin. Figure S9. High-grade serous ovariancancer Patient 259 derived CSC drive elevated CD206 expression in mac-rophages, and demonstrate a carboplatin resistant phenotype. FigureS10. Scatter plots for correlation of WNT5B with immune cell subsets inovarian carcinoma. Table S1. List of primers used for qPCR experiments.(ZIP 1916 kb)

AcknowledgementsThe authors also acknowledge the efforts of Dr. Joel Whitfield from theImmunological Monitoring Core at the Rogel Cancer Center, for performingELISA measurements. The authors acknowledge the efforts of Mr. MarkSavary at the Flow Cytometry Core for technical assistance in performingfluorescent activated cell sorting. The authors also acknowledge Mr. NamHoon Kim, supported by the Michigan LS&A Undergraduate ResearchOpportunity Program, for experimental support. Lastly, the authorsacknowledge the services of the University of Michigan Cancer Center Tissue

Core, for performing routine histology and immunohistochemistry tosupport this manuscript.

Authors’ contributionsS.R. designed and performed experiments, interpreted data, and wrote themanuscript. P.M. performed and interpreted in vivo studies, and edited themanuscript. Y.X. performed bioinformatic analysis. Y.L.L designed andinterpreted studies, provided specimen, edited the manuscript and providedfunding. G.M. conceptualized the study, designed the experiments,interpreted the data, wrote the manuscript, and provided tissue specimensand funding. All authors read and approved the final manuscript.

FundingThis work was supported by the DOD OCRP W81XWH-13-1-0134 (GM), Mich-igan Ovarian Cancer Alliance (MIOCA, GM), Rivkin Center for Ovarian Cancer(GM), NIH/NIDCR DE026728 (YLL) and the Tissue Engineering and Regener-ation Training Grant under award number NIH/NIDCR T32DE00007057 (SR).Flow cytometry facilities used to perform experiments on this manuscriptwere partially funded by NIH/NCI P30CA046592. The funding agencies hadno roles in the design of the study and collection, analysis, and interpretationof data and in writing the manuscript.

Availability of data and materialsPlease contact corresponding author for data requests.

Ethics approval and consent to participateNo human subjects.All animal experiments were conducted with approval of the UniversityCommittee on Use and Care of Animals at the University of Michigan.

Consent for publicationNot applicable.

Competing interestsThe authors declare that they have no competing interests.

Author details1Department of Materials Science and Engineering, 2800 Plymouth Rd,Building 28, Room 3044W, Ann Arbor, MI 48109, USA. 2Department ofBiomedical Engineering, 2800 Plymouth Rd, Building 28, Room 3044W, AnnArbor, MI 48109, USA. 3Department of Macromolecular Sciences andEngineering, 2800 Plymouth Rd, Building 28, Room 3044W, Ann Arbor, MI48109, USA. 4Department of Periodontics and Oral Medicine and Departmentof Otolaryngology Head and Neck Surgery, Ann Arbor, USA. 5Rogel CancerCenter, North Campus Research Complex, University of Michigan, 2800Plymouth Rd, Building 28, Room 3044W, Ann Arbor, MI 48109, USA.6Department of Computational Mathematics, Science, and Engineering,Michigan State University, East Lansing, MI 48823, USA.

Received: 14 February 2019 Accepted: 7 July 2019

References1. Liao J, Qian F, Tchabo N, et al. Ovarian cancer spheroid cells with stem cell-

like properties contribute to tumor generation, metastasis andchemotherapy resistance through hypoxia-resistant metabolism. PLoS One.2014;9:e84941.

2. Shield K, Ackland ML, Ahmed N, Rice GE. Multicellular spheroids in ovariancancer metastases: biology and pathology. Gynecol Oncol. 2009;113:143–8.

3. Allen HJ, Porter C, Gamarra M, Piver MS, Johnson EA. Isolation andmorphologic characterization of human ovarian carcinoma cell clusterspresent in effusions. Exp Cell Biol. 1987;55:194–208.

4. Kipps E, Tan DS, Kaye SB. Meeting the challenge of ascites in ovarian cancer:new avenues for therapy and research. Nat Rev Cancer. 2013;13:273–82.

5. Zhang L, Conejo-Garcia JR, Katsaros D, et al. Intratumoral T cells, recurrence,and survival in epithelial ovarian cancer. N Engl J Med. 2003;348:203–13.

6. Hamanishi J, Mandai M, Iwasaki M, et al. Programmed cell death 1 ligand 1and tumor-infiltrating CD8+ T lymphocytes are prognostic factors of humanovarian cancer. Proc Natl Acad Sci U S A. 2007;104:3360–5.




j.com/

J Imm


ownloaded from

https://doi.org/10.1186/s40425-019-0666-1


7. Yin M, Li X, Tan S, et al. Tumor-associated macrophages drive spheroidformation during early transcoelomic metastasis of ovarian cancer. J ClinInvest. 2016;126:4157–73.

8. Lan C, Huang X, Lin S, et al. Expression of M2-polarized macrophages isassociated with poor prognosis for advanced epithelial ovarian cancer.Technol Cancer Res Treat. 2013;12:259–67.

9. Hagemann T, Wilson J, Burke F, et al. Ovarian cancer cells polarize macrophagestoward a tumor-associated phenotype. J Immunol. 2006;176:5023–32.

10. Zhang M, He Y, Sun X, et al. A high M1/M2 ratio of tumor-associatedmacrophages is associated with extended survival in ovarian cancerpatients. J Ovarian Res. 2014;7:19.

11. Reinartz S, Schumann T, Finkernagel F, et al. Mixed-polarization phenotypeof ascites-associated macrophages in human ovarian carcinoma: correlationof CD163 expression, cytokine levels and early relapse. Int J Cancer. 2014;134:32–42.

12. Yafei Z, Jun G, Guolan G. Correlation between macrophage infiltration andprognosis of ovarian cancer- a preliminary study. Biomed Res. 2016;27:305.

13. Raghavan S, Mehta P, Ward MW, et al. Personalized Medicine BasedApproach to Model Patterns of Chemoresistance and Tumor RecurrenceUsing Ovarian Cancer Stem Cell Spheroids. Clin Cancer Res. 2017.

14. Raghavan S, Ward MR, Rowley KR, et al. Formation of stable small cellnumber three-dimensional ovarian cancer spheroids using hanging droparrays for preclinical drug sensitivity assays. Gynecol Oncol. 2015;138:181–9.

15. Raghavan S, Mehta P, Horst EN, Ward MR, Rowley KR, Mehta G. Comparativeanalysis of tumor spheroid generation techniques for differential in vitrodrug toxicity. Oncotarget. 2016.

16. Barbolina MV, Burkhalter RJ, Stack MS. Diverse mechanisms for activation ofWnt signalling in the ovarian tumour microenvironment. Biochem J. 2011;437:1–12.

17. Fodde R, Brabletz T. Wnt/beta-catenin signaling in cancer stemness andmalignant behavior. Curr Opin Cell Biol. 2007;19:150–8.

18. Spranger S, Bao R, Gajewski TF. Melanoma-intrinsic beta-catenin signallingprevents anti-tumour immunity. Nature. 2015;523:231–5.

19. Anagnostou SH, Shepherd PR. Glucose induces an autocrine activation ofthe Wnt/beta-catenin pathway in macrophage cell lines. Biochem J. 2008;416:211–8.

20. Feng Y, Ren J, Gui Y, et al. Wnt/beta-catenin-promoted macrophagealternative activation contributes to kidney fibrosis. J Am Soc Nephrol. 2018;29:182–93.

21. Pukrop T, Klemm F, Hagemann T, et al. Wnt 5a signaling is critical formacrophage-induced invasion of breast cancer cell lines. Proc Natl Acad SciU S A. 2006;103:5454–9.

22. Ojalvo LS, Whittaker CA, Condeelis JS, Pollard JW. Gene expression analysisof macrophages that facilitate tumor invasion supports a role for Wnt-signaling in mediating their activity in primary mammary tumors. JImmunol. 2010;184:702–12.

23. Wang X, Deavers M, Patenia R, et al. Monocyte/macrophage and T-cellinfiltrates in peritoneum of patients with ovarian cancer or benign pelvicdisease. J Transl Med. 2006;4:30.

24. Balkwill FR, Capasso M, Hagemann T. The tumor microenvironment at aglance. J Cell Sci. 2012;125:5591–6.

25. Cortes M, Sanchez-Moral L, de Barrios O, et al. Tumor-associatedmacrophages (TAMs) depend on ZEB1 for their cancer-promoting roles.EMBO J. 2017;36:3336–55.

26. Chimal-Ramirez GK, Espinoza-Sanchez NA, Chavez-Sanchez L, Arriaga-PizanoL, Fuentes-Panana EM. Monocyte differentiation towards Protumor activitydoes not correlate with M1 or M2 phenotypes. J Immunol Res. 2016;2016:6031486.

27. Shen L, Li H, Shi Y, et al. M2 tumour-associated macrophages contribute totumour progression via legumain remodelling the extracellular matrix indiffuse large B cell lymphoma. Sci Rep. 2016;6:30347.

28. Sharp BM. Conversion of the U937 Monocyte into" Macrophage-Like"Populations Exhibiting M1 or M2 Characteristics2013.

29. Gupta V, Yull F, Khabele D. Bipolar Tumor-Associated Macrophages inOvarian Cancer as Targets for Therapy. Cancers (Basel). 2018;10.

30. Ko SY, Ladanyi A, Lengyel E, Naora H. Expression of the homeobox geneHOXA9 in ovarian cancer induces peritoneal macrophages to acquire an M2tumor-promoting phenotype. Am J Pathol. 2014;184:271–81.

31. Chuang Y, Hung ME, Cangelose BK, Leonard JN. Regulation of the IL-10-driven macrophage phenotype under incoherent stimuli. Innate Immun.2016;22:647–57.

32. Makita N, Hizukuri Y, Yamashiro K, Murakawa M, Hayashi Y. IL-10 enhancesthe phenotype of M2 macrophages induced by IL-4 and confers the abilityto increase eosinophil migration. Int Immunol. 2015;27:131–41.

33. Deng X, Zhang P, Liang T, Deng S, Chen X, Zhu L. Ovarian cancer stem cellsinduce the M2 polarization of macrophages through the PPARgamma andNF-kappaB pathways. Int J Mol Med. 2015;36:449–54.

34. Kokubu Y, Tabu K, Muramatsu N, et al. Induction of protumoral CD11c(high)macrophages by glioma cancer stem cells through GM-CSF. Genes Cells.2016;21:241–51.

35. Fan QM, Jing YY, Yu GF, et al. Tumor-associated macrophages promotecancer stem cell-like properties via transforming growth factor-beta1-induced epithelial-mesenchymal transition in hepatocellular carcinoma.Cancer Lett. 2014;352:160–8.

36. Lu H, Clauser KR, Tam WL, et al. A breast cancer stem cell niche supportedby juxtacrine signalling from monocytes and macrophages. Nat Cell Biol.2014;16:1105–17.

37. Arend RC, Londono-Joshi AI, Straughn JM Jr, Buchsbaum DJ. The Wnt/beta-catenin pathway in ovarian cancer: a review. Gynecol Oncol. 2013;131:772–9.

38. Katoh M. Canonical and non-canonical WNT signaling in cancer stem cellsand their niches: cellular heterogeneity, omics reprogramming, targetedtherapy and tumor plasticity (review). Int J Oncol. 2017;51:1357–69.

39. Takebe N, Harris PJ, Warren RQ, Ivy SP. Targeting cancer stem cells by inhibitingWnt, notch, and hedgehog pathways. Nat Rev Clin Oncol. 2011;8:97–106.

40. Schaale K, Brandenburg J, Kispert A, Leitges M, Ehlers S, Reiling N. Wnt6 isexpressed in granulomatous lesions of Mycobacterium tuberculosis-infectedmice and is involved in macrophage differentiation and proliferation. JImmunol. 2013;191:5182–95.

41. Derlindati E, Dei Cas A, Montanini B, et al. Transcriptomic analysis of humanpolarized macrophages: more than one role of alternative activation? PLoSOne. 2015;10:e0119751.

42. Ferraro NM, Dampier W, Weingarten MS, Spiller KL. Deconvolution ofheterogeneous wound tissue samples into relative macrophage phenotypecomposition via models based on gene expression. Integr Biol (Camb).2017;9: 328–338.

43. van Dijk EM, Menzen MH, Spanjer AI, Middag LD, Brandsma CA, Gosens R.Noncanonical WNT-5B signaling induces inflammatory responses in humanlung fibroblasts. Am J Physiol Lung Cell Mol Physiol. 2016;310:L1166–76.

44. Christodoulides C, Lagathu C, Sethi JK, Vidal-Puig A. Adipogenesis and WNTsignalling. Trends Endocrinol Metab. 2009;20:16–24.

45. Kato S, Hayakawa Y, Sakurai H, Saiki I, Yokoyama S. Mesenchymal-transitioned cancer cells instigate the invasion of epithelial cancer cellsthrough secretion of WNT3 and WNT5B. Cancer Sci. 2014;105:281–9.

46. Harada T, Yamamoto H, Kishida S, et al. Wnt5b-associated exosomespromote cancer cell migration and proliferation. Cancer Sci. 2017;108:42–52.

47. Smith K, Bui TD, Poulsom R, Kaklamanis L, Williams G, Harris AL. Up-regulation of macrophage wnt gene expression in adenoma-carcinomaprogression of human colorectal cancer. Br J Cancer. 1999;81:496–502.

48. Kaler P, Godasi BN, Augenlicht L, Klampfer L. The NF-kappaB/AKT-dependent induction of Wnt signaling in Colon Cancer cells bymacrophages and IL-1beta. Cancer Microenviron. 2009;2:69–80.

49. Cosin-Roger J, Ortiz-Masia D, Calatayud S, et al. M2 macrophages activateWNT signaling pathway in epithelial cells: relevance in ulcerative colitis.PLoS One. 2013;8:e78128.

50. Gattinoni L, Ji Y, Restifo NP. Wnt/beta-catenin signaling in T-cell immunityand cancer immunotherapy. Clin Cancer Res. 2010;16:4695–701.

51. Wellenstein MD, de Visser KE. Cancer-cell-intrinsic mechanisms shaping thetumor immune landscape. Immunity. 2018;48:399–416.

52. Galluzzi L, Spranger S, Fuchs E, Lopez-Soto A. WNT signaling in CancerImmunosurveillance. Trends Cell Biol. 2019;29:44–65.

53. Gattinoni L, Zhong XS, Palmer DC, et al. Wnt signaling arrests effector T celldifferentiation and generates CD8+ memory stem cells. Nat Med. 2009;15:808–13.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.




j.com/

J Imm


ownloaded from


ovarian cancer stem cells and macrophages reciprocally ...educated by tumor-derived factors into an...

Documents